How can quantitative strategies be classified? Taxonomy is an essential part of science and an important, albeit sometimes overlooked part of finance. Even in data science, the core activity of unsupervised learning is to form some classification system. Quant strategies come in all types and use a multitude of data science techniques for classification; nevertheless, the key to understanding strategies is through classification. Unfortunately, many quant managers do not want to be typecast through a rigorous system because it may diminish their perceived alpha. Classification, however, allows for skill measurement and some form of prediction for when a strategy will work and when it will underperform.
Most of quant strategies can be classified as factor-based. The manager will make investment decisions based on a single factor like momentum or value. There value-added is through the method used to extract that factor. One value manager may be different than another through their system or process of finding and exploiting the risk factor. However, the single factor can serve as the best explanation of return variation.
Since risk factors can apply to any asset class, the taxonomy will classify the quant by the factor or set of factors that dominate the risk exposure and the asset class employed. A quant manager could be an equity value, credit carry, or multi-asset momentum. The risk factor must be well-defined and well-known to serve as a factor descriptor. The large factor zoo that can explain return variation explodes the taxonomy but still makes it informative.
An enhanced version of factor investing is when a quant uses some data set to dynamically adjust or change multiple factor weights. Factors are measured at any specific time and adjusted or constrained by a second set of variables such as the business cycle. We can call this dynamic factor weighting.
Within any quant strategy there can be further classification based on the form of implementation. This will be the portfolio construction. For example, a risk parity approach or volatility control can serve as a tertiary rule for classification.
The second set of quant strategies are those that cannot be explained by a well-defined factor and hence can be classified as an alpha strategy. These strategies are unique from any well-known factor and hard to classify and thus must be given a verbal description. The inability to classify makes them extremely useful because they may have the strongest diversification benefits. Unfortunately, while it may prove to be useful to form a strategy grouping by alpha, it may be hard to predict future returns other than through extrapolation.
Classification systems for quants or any hedge funds by factors provide a more detailed and objective measure of fund behavior than any approach that is based on self-identification or description and thus helps to truly identify skill.
No comments:
Post a Comment